Extend cuda_flip to unsigned types (pytorch#137781)

Using AT_DISPATCH_V2

Test plan: `python3 -c "import torch;print(torch.randint(0, 100, (4, 4),  dtype=torch.uint16, device='cuda').flip(0))"`
Fixes pytorch#137770

Pull Request resolved: pytorch#137781
Approved by: https://github.com/Skylion007

Author: malfet

aten/src/ATen/native/cuda/IndexKernel.cu

@@ -457,12 +457,20 @@ void flip_kernel(TensorIterator& iter, const bool quantized) {
       flip_kernel_impl<dtype>(iter);
     });
   } else {
-    AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND3(at::ScalarType::Half, at::ScalarType::Bool, at::ScalarType::BFloat16,
-                                           iter.dtype(), "flip_cuda",
-    [&] {
-      using dtype = OpaqueType<sizeof(scalar_t)>;
-      flip_kernel_impl<dtype>(iter);
-    });
+    AT_DISPATCH_V2(
+      iter.dtype(),
+      "flip_cuda",
+      AT_WRAP([&] {
+        using dtype = OpaqueType<sizeof(scalar_t)>;
+        flip_kernel_impl<dtype>(iter);
+      }),
+      AT_EXPAND(AT_ALL_TYPES_AND_COMPLEX),
+      AT_EXPAND(AT_FLOAT8_TYPES),
+      AT_EXPAND(AT_BAREBONES_UNSIGNED_TYPES),
+      kComplexHalf,
+      kHalf,
+      kBool,
+      kBFloat16);
   }
 }